Maker tests GPT-6 Astra for mechanical design of a latte art camera

The initial prototype failed because the button mechanism tilted and snagged due to uneven force distribution, a flaw not visible in ideal CAD animations. The AI analyzed photos of the malfunction and proposed three structural fixes. The maker selected a lever-based solution to redirect force to the center, which resolved the issue. This process highlighted that while AI can generate 3D shapes, it requires human feedback on physical constraints like friction and deflection to create functional mechanisms.

The project demonstrates a shift in CAD workflows where AI handles geometry and iteration while humans focus on design intent and physical verification. The author argues that "being able to create a 3D model" is distinct from "designing something that actually works," and that AI can assist in diagnosing root causes of mechanical failures when provided with real-world evidence.